TWI239207B - Method and device for coding digital images using intra-mode block prediction - Google Patents

Abstract

A method and device for coding digital images using intra-mode block prediction, wherein a list of prediction modes for each combination of prediction modes of the neighboring blocks is obtained. The modes assigned to each combination of prediction modes may be divided into two groups. The first group includes n (where n is smaller than the overall number of available modes) most probable prediction modes and the second group includes the remaining modes. The modes in the first group are ordered according to their probability. This order may be specified as a list of modes ordered from most probable to the least probable mode. The modes belonging to the second group may be ordered in some predetermined manner, which may be specified depending on the information already available to the decoder.

Description

说明 Description of the invention: [Technical field to which the invention belongs] The present invention relates to the encoding of images, and more specifically to the encoding of data block groups of video frames. [Previous technology] Pigs such as video images, TV images, πI color images are composed of video and digital images produced by video cameras or% monthly, and are composed of pixels arranged in horizontal and vertical lines. The number of pixels in a single image is usually ten thousandths. Each pixel generally contains luminance and chrominance information. Without compression, the amount of information transmitted from the-image encoder to another-the image encoder is so secretive that real-time image transmission cannot be performed. In order to reduce the amount of information transmitted, several different compression methods have been developed, including JPEG and the Η 263 standard. In a typical video encoder, the frame of the starting video sequence is divided into rectangular areas or data blocks: material type (1, type) or intermediate mode (ρ-mode) for encoding. Some kind of transformation ... The DCT encoding method will be used in the block: ,,, and the encoding of pure pure Beco blocks can only reduce specific data == inter-pixel correlation, and will not consider the intermediate data of pixels & ghost correlation 'Bit image coding standards that still produce high bit rates also reveal specific ways to reduce the correlation. Phases of pixel values between κ and ^^ Middle == coded data blocks can be encoded from the previous and produced (2D) n. The predictive information of the data block is given by 1) the motion vector ... "Code 2 under chess: Chinese 'indicates that the data block is formed by using the spatial prediction in the neighboring data block that has been coded in the same frame. The prediction error is the coded data block and the predicted data block. The difference between them is represented by a set of specificity functions of discrete transformations. The transformations are generally performed on 8x 8 or 4x 4 data block benchmarks. Gravity—transformation coefficients will be quantified later. Quantification will introduce information Loss, so the quantized coefficients have lower accuracy than the original ones. The quantized transform coefficients together with the motion vector and some control information will form a complete coding sequence representation, represented by syntax elements. It is transmitted from the encoder to the decoder. Previously, all syntax elements were entropy-coded to further reduce the number of bits required for representation. In the decoder, the data blocks in the current frame are taken from the first prediction of the same way as the blood encoder, and will be compressed The prediction error is added to the prediction. The compression coefficient is used to obtain the proportion of the transform function and the compression prediction & difference is found. The difference between the reconstructed frame and the original frame The difference between them is called reconstruction error. The compression ratio, which is the number of bits used to represent the original and compression order, and the coefficients and data blocks, is controlled by adjusting the value of the two used to quantify. The compression ratio also depends on the entropy encoding method used. Jvt encoder> a # + i Draft number 2 (WD2) The space used in advance provides 4 X 4 data blocks. 'J VVT encoder 0) A 8 ^ 9 kinds of predictive modes, including DC predictive (mode type) as shown in the figure

The pixels of p will be Q pixels.

The procedure is shown in Figure 2. Referring to Fig. 2, from a to p are coded, and A to Q of adjacent data blocks that have been coded are used as foresights. For example, when the mode 1 is selected, the pixels a, e, i, and are set to be equal to the pixel A to be predicted, and the pixels b, f,], and are set to be equal to the pixel B to be predicted. Similarly, if mode 2 is selected, pixels a, b, c, and d are set equal to pixels! It is predicted that the pixels e, f, g, and h are set to be equal to the pixel j and predicted. Therefore mode i is a predictor in the vertical direction; mode 2 is a predictor in the horizontal direction. This model is described in the VCEG-N54 document published by the Video Coding Experts Group (VCEG) 1, the ITU-11 Telecommunication Standardization Sector, published in September 2001, and described in March 2002. In the JVT_B118r2 file published by the joint video group of ISO / IEC MPEG and ITU-T VCEG. Mode 0: DC pre-recording Generally, all samples are predicted using (a + b + c + d + i + j + k + l + 4) > > 3. If four of the samples exceed the image, the average of the remaining four is used for prediction. If all eight samples exceed the image, the prediction for all samples in the data block is 128. Therefore, this mode can be used to prevent data blocks. Model L: Hanging Mu? § shows that if A, B, C, and D are within the image range, then-a, e, i, m are foretold by A, -b, f, j, and n are foreshadowed by β,-c, g , K, 0 is foretold by c, -d, h, 1, p is foretold by D. Mode 2: Horizontal indicates that if E, F, G, and Η are within the image range, J -a, b, c, and d are indicated by E, and -e, f, g, and h are indicated by F, -i, j, k, 1 are predicted by G, -m, η, ο, p are predicted by Η. Mode 3: Opposite Down / Right Prediction This mode is only used when A, B, C, D, I, J, K, L, and Q are in the image range. This is a π diagonal π prediction. -m is predicted by (J + 2K + L + 2) > &2; -i, η is predicted by (I + 2J + K + 2) > &2; -e, j, 〇 is predicted by (Q + 2I + J + 2) > 2, -a, f, k, p is predicted by (A + 2Q + I + 2) > 2, -b, g, 1 is predicted by (Q + 2A + B + 2) > &2; -c, h is predicted by (A + 2B + C + 2) > &2; -d is caused by (B + 2C + D + 2) > > Mode 4: Downward / Left Prediction for Yu This mode is only used when all A, B, C, D, I, J, K, L, and Q are within the image range. -a is determined by (A + 2B + C + I + 2J + K + 4) > > 3 '-b, e is determined by (B + 2C + D + J + 2K + L + 4) > > 3 does not predict '-c, f, i are given by (C + 2D + E + K + 2L + M + 4) > > 3, -d, g, j, m are given by ( D + 2E + F + L + 2M ten N + 4) > > 3

-h, k, η is predicted by (E + 2F + G + M + 2N + 0 + 4) > > 3, -1, o is determined by (F + 2G + H + N + 20 + P + 4) > > 3, and -p is predicted by (G + H + 0 + P + 2) > > 3. Mode 5: Vertical-Left Prediction This mode is only used when all A, B, C, D, I, J, K, L, Q are in the image range. This is the π diagonal π prediction. -a, j is foretold by (Q + A + 1) > &1;-b, k is foretold by (A + B + 1) > > 1, -c, 1 is from (B + C + 1) > > 1 indicates that -d is indicated by (C + D + 1) > > 1, -e and η are indicated by (I + 2Q + A + 2) > > 2 indicates that -f, 〇 is indicated by (Q + 2A + B + 2) > > 2, -g, p is indicated by (A + 2B + C + 2) > > 2 , -H is predicted by (B + 2C + D + 2) > > 2, -i is predicted by (Q + 2I + J + 2) > > 2, -m is predicted by (I + 2J + K + 2) > > 2 predicted. Mode 6: Vertical-right indicates that this mode is only used when all A, B, C, D, I, J, K, L, and Q are in the image range. This is π diagonal. ”-A (2A + 2B + J + 2K + L + 4)> 3 indicates that-b, i is indicated by (B + C + 1) > &1; -c, j is indicated by (C + D +1) > > 1,-d, k is foreshadowed by (D + E + 1) > > 1, -1 is foreshadowed by (E + F + 1) > > 1 , 9 -e is indicated by (A + 2B + C + K + 2L + M + 4) > > 3, -f, in is indicated by (B + 2C + D + 2) > > 2 No, -g, η is predicted by (C + 2D + E + 2) > > 2, -h, 〇 is predicted by (D + 2E + F + 2) > > 2,- p is foreshadowed by (E + 2F + G + 2) > > 2. Mode 7: Level-up indicates that this mode is only used when all A, B, C, D, I, J, K, L, Q is in the range of the image, and this system is π diagonal. -a is foretold by (A + 2C + D + 2I + 2J + 4) > > 3, -b is for (C + 2D + E + I + 2J + K + 4) > > 3 Foreshadow, -c 'e is predicted by (D + 2E + F + 2J + K + 4) > > 3, -d, f is predicted by (E + 2F + G + J + 2K + L + 4 ) > > 3 indicates that -g, i is indicated by (F + 2G + H + 2K + 2L + 4) > > 3 indicates that -h, j is indicated by (G + 3H + K + 3L +4) > > 3 indicates that -1, η is indicated by (L + 2M + N + 2) > > 3 indicates that -k, m is indicated by (G + H + L + M + 2 ) >> 2, -0 is predicted by (M + N + 1) > > 1, -p is predicted by (M + 2N + 0 + 2) > > 2. Mode 8: Horizontal-Down Prediction This mode is only used when all A, B, C, D, I, J, K, L, and Q are within the image range. -a, g is predicted by (Q + I + 1) > > 1, -b, h is predicted by (I + 2Q + A + 2) > > 2, -c is expressed by (Q + 2A + B + 2) > > 2 indicates that 10

& lj-d is predicted by (A + 2B + C + 2) > > 2 '-e, k is predicted by (I + J + 1) > > 1, -f, 1 Is indicated by (Q + 2I + J + 2) > > 2, -i, 〇 is indicated by (J + K + 1) > > 1, -j, P is indicated by (1 + 2 J + K + 2) > > 2 indicates that -m is indicated by (K + L + 1) > > 1,-η is indicated by (J + 2K + L + 2) > > 2 foreshadowed. Since each data block must have a predictive mode specified and transmitted to the decoder, a significant number of bits are required for direct encoding. In order to reduce the information to be transmitted, the interrelationship of the predictive mode of adjacent data blocks can be used. For example, Vahteri et al. (WO001 / 54416A1, "Methods for Encoding Images and Image Encoders", hereinafter represented by Vahteri) disclose a coding method based on data blocks, in which the The directional information of the image will be used to classify multiple spatial predictive modes. The spatial prediction pattern of the body block depends on the directional information of at least one adjacent data block. In the JVT encoder, when the adjacent predictive mode is known to be the encoded data blocks U and L, the order of the highest probability predictive mode of data block c, the next highest probability predictive mode, etc. (Figure 3) . The order of the modes that indicate the various combinations of modes U and L is specified. This order is referred to as the list of predictive patterns of data block C, which are ranked from the highest probability to the lowest probability. As revealed in vceg-N54, the sequence list used in jvt editing WD2 is as follows:

Table I: Predictive mode using a sorted signal as a function in the bit stream L / U outside 0 1 2 3 outside -------- 0 -------- 01 ------- 10 ------- 0 02 ------- 021648573 125630487 021876543 021358647 1 102654387 162530487 120657483 102536487 2 20 ------- 280174365 217683504 287106435 281035764 3 201385476 125368470 208137546 325814670 4 201467835 162045873 204178635 420615837 5 015263847 152638407 201584673 531286407 6 016247583 160245738 206147853 160245837 7 270148635 217608543 278105463 270154863 8 280173456 127834560 287104365 283510764 L / U 4 5 6 7 8 outside -------- 0 206147583 512368047 162054378 204761853 120 208 685 685 1685 215368740 216748530 278016435 287103654 3 421068357 531268470 216584307 240831765 832510476 4 426015783 162458037 641205783 427061853 204851763 5 125063478 513620847 165230487 210856743 210853647 6 640127538 165204378 614027538 264170583 615 853 083 083 685 6153 8 287461350 251368407 216847350 287410365 283074165 As specified in WD2 of the JVT encoder, when the predictive mode of U and L is 2, an example of the predictive mode of data block C will be provided. The number (2, 8, 7 small 0, 6, block C, the highest probability mode, 3, 5) represents that mode 2 is also the data area, and mode 8 is the mode with the second lowest probability, and so on 'and so on. For the decoder, the information to be transmitted represents the n-th highest probability mode that data block C will use. The ordering of the mode of data block C can also be specified by the ranking of various modes. The higher the ranking, the lower the probability of the predictive method. In the above embodiment, the ranking would be (5, 4, 12

2) Hour ... ,, y, b, 3, 2) 〇 When the pattern 7'8) is related to the ranking list (5,4, Bu 8,7,9,6, 'It can be seen that the pattern 0 ranks 5, pattern The ranking of ^ is 4 and so on. — For more efficient coding, the information in two 4 × 4 data blocks can be encoded into a code character. The above method has a major disadvantage, maintaining the data block. The ordering of the predictive mode to provide the predictive mode of the data block requires a large amount of memory. In the WD2 of the m encoder, there are 9 × 9 possible combinations of modes that predict the 'data block' by using 9 modes. Various The combination must specify the ordering of 9 possible modes. In other words, 9 × 9x_ 9 bits are required (assuming here—the number needs—the number of bits) to specify the ordering of the pre-patterns. In addition, more memory is needed to specify Special cases, such as when one or two data blocks M l do not exist. Therefore, it is better to provide a method and device for encoding digital images. The memory requirement is reduced and the loss of encoding efficiency is the lowest. "[Summary of the invention] The present invention provides a use Method and device for pattern predictive data block encoding digital image. A list of various combinations of predictive patterns adjacent to the data block can be obtained. The patterns of various combinations assigned to the predictive patterns can be divided into two groups. The first group includes m highest Probabilistic predictive patterns (where m is less than the total number of existing patterns η), and the second group includes the remaining modulo 13

弋 Group—The patterns are sorted according to their probability. This order can be deducted as a sorted list of the most probable to the least probable mode. Belonging to and can be sorted in a certain preset way, and can be specified according to the I news that exists in Xie Majie. Information about the mode selected for a particular data block belongs to the first or second group will be transmitted to the decoder. ^ The fruit belongs to the-group, and the information will be transmitted to indicate the i-th highest machine = mode applied to the data block α for the mode combination of the data block UAL. If the mode belongs to the second group, the information will be transmitted indicating that the jth mode of this group should be used. The invention will become clearer with reference to the detailed description of Figs. 4a to 8. . [Embodiment] An embodiment of the present invention uses the predictive mode ^ ^ combination of adjacent data blocks as a function of the predictive mode of another combination to obtain the special order of Γ ^. For illustrative purposes, the predictive patterns of the Nitian neighbor data blocks ϋ and L shown in Figure 3 are used to infer the prediction of the existing data block c. The combination of the predictive patterns in Figure 4a can be made by diagonally flipping the cooperation of the predictive patterns in the flap. Therefore, when the 榲 combination in Fig. 4a is used, the n-th highest probability of data block c indicates the same as the n-th highest probability of the pattern combination in Fig. 4b: The angle flipper is the same. Therefore, if the adjacent data block ϋ and L have, "G" and "vertical" mode, the current predictive mode of the existing data block c is the most honey (C 4b). As a result, when the data block is flipped or diagonal shovel

Shoot (" downward / rightward "), but you can get the" level "of the existing data block" from the water level and level "(Section 4a sm m. Private picture). Similarly, if the adjacent data blocks ϋ and L · are the grips 4 9 and 8 Z and 3 as shown in Fig. 5a, the data blocks u and L will be turned as patterns 3 and i shown in Fig. 5b. ^ ^ ^ 詋 In this embodiment, the predictive direction i is mapped to j: .fj Η1), the meaning is as follows: each predictive mode i takes a predictive mode J $ 预 τ 板式 j is composed along the diagonal The upper left corner of the data block is mirrored to the lower right corner of the person in the data area. The final orientation of the prophetic model in Figure 1 is summarized below.

Table Π, when describing the function of the mode, the predictive mode group of the pattern (k, 1) is based on Zheng Zheng 'in the cascade (... ·), that is, if the (n, nth highest probability embodiment of the combination (i, D) is considered, the data model Is equal to f (p) e as real

The pattern of the list of the pattern of L (丨, D 15 uses two, ::: (22: 75 ::: °, 4'8, 7). The list can be sorted from 8 of the predictive pattern of (2, 2) 'Second, :: 4, 6, 5) maps to get 7 and the sorted list of V \ f (2' 3) = (3, 丨) ...,. , 8, i, 6 ;, The mapping ^ sorted list is roughly for ⑽ (i, j). Therefore, the shot f can be interpreted as a kind of mirror function. The main purpose of Ming is to reduce the size of the predictive model that has been edited. The "Beijing block (for example, the list of the predictive model of the function of Table 0) can be obtained during the internship process. To obtain the block of late neighbor data :,: A part of the combined data block C with a larger number of patterns (i, D ^ pre-7F mode sorted list, the data of each predictive mode is known and selected and selected (such as the minimum predictive error) for calculation. It is determined that the predictive mode is assigned to the combination 0,]). The predictive mode β is the ranking of the predictive mode in the early days. The more frequently the mode is selected, the higher the ranking in the list of modes. The predictive mode will not be used during the above internship. When mapping, the generated list of pre-patterns will be smaller. If the sorted list of combination (k, 1) can be obtained from the sorted list of combination (i, j), only the combination (i, "·) The sorting list needs to be stored. During the internship, the mapping operation can be performed as follows. Each of the combinations (k, n and (i, J ·) i = f (0 and j = f (k) indicates the frequency of 16 patterns. Is a joint calculation, that is, if the predictive mode P is selected For combination (k, 1), it is used as the predictive mode f (p) for combination (i, j). When the predictive mode s is selected for combination (i, j), it is regarded as the predictive mode f (s). It is selected for combination (k, 1). The reduction foresight of the present invention is shown in the following table: Table II:] Λ Reduction foresight Table 1 2 η * ι 1 π 3 024167835 150642387 027486135 013245867 150643278 021468735 105436287 124086573 283407156 385240167 L / U outside outside- ------ 0 02 1 — 2 20 3-4 —— 5-6-7 — 8 —— L / U 4 5 6 7 8 outside 0 012465738 150346287 160452387 024716835 028413765 1 104562378 156403287 165403278 014652733 014256837 2 240781635 214835076 241086735 207483165 280473156 3 413205876 531480267 146530287 247308516 832045176 4 420671835 145602387 461027538 407261835 248073165 5 513406287 165402387 240158376 082354 703U 682 A sorted list of models is not provided. The sorted list of the combination can be retained in the preview table by mapping on page 17

El should always "reform" the "reform", which is required for the prediction of the current Q-block. So in general, as long as the elements in the = table can be obtained or the other elements in the predictive table are passed through the :: mapping using the mapping: the former can be eliminated. In other words, in the prediction table with the first group of elements, each of the second group of elements can be reformed from the corresponding element in the first group by using the beam shooting function, and the second group of elements can be eliminated . Figure 6 is a flowchart of the decoding phase when the symmetry in the predictive table is used. As shown in the figure, the step of method 1) includes receiving a plurality of image data blocks in step 110. #When the existing data block is processed, step 120 determines whether the predictive mode of the existing data block can be obtained from the predictive mode of the E-lin data block without custom mapping. If possible, the spatial prediction mode of the existing data block may depend on the prediction mode of the adjacent data block in step 132. Otherwise, it will be mentioned in step] [30: the complementary predictive mode of the existing data block, and the complementary predictive mode of the existing data block will be determined in step 14 according to the complementary predictive mode of the Tianbi neighbor tributary block. In step 150, the complementary predictive mode of the existing data block is mapped into the predictive mode of the existing data block. . On the other hand, before using data blocks ϋ and L to define the predictive mode of data block c, the same indication can be indicated to the data block μ L = the same predictive mode (grouping them together). For example, in the case of a JVT encoder, modes 1 > 5 and 6 can be divided into a group and labeled as Bu while nuclear formulas 2, 7 and 8 can be divided into a group and labeled as 2. As shown in Figure i, the directions of Modes 7 and 8 are close to Mode 2, and the directions of Modes 5 18! Μ Βί and 6 are close to Mode 1. After this grouping, each data block U and L has one of five patterns labeled 0, 1, 2, 3, and 4. Therefore, from the original 9 × 9 possible combinations of the U and L predictive patterns to only 5 × 5 combinations. Therefore, under the predictive mode of the known data block U, the memory required to sort the predictive mode of the data block C will be changed from the original 9 X 9 X 9 bits to 5 X 5 X 9 bits Yuan (assuming that one bit requires a bit of memory). In addition, if the mapping function f is used to '' flip the π sorted list, the predictive list can be further simplified. When the two methods are jointly adopted, one embodiment of the table specifying the predictive mode as a function of the ordering number in the bit stream is shown in Table IV.

Table IV L / U outside

Outside ------- 0 02 ------- 1 --------- 2 20 ------- 3 --------- 4 --- ------ 0 1 2 3 4 0 -------- 01 ------- 10 ------- 024167835 150642387 024781635 013245867 012465738 156043278 021468375 153046827 140652378 214806357 283407156 247081635 385240167 413205876 420671835 Furthermore, the number of predictive modes of data block C can also be limited under known predictive modes of data blocks U and L. In the case of JVT encoders, there are still 9x 9 possible combinations of U and L predictive modes. However, each combination can only specify m modes, where m is less than 9. The number of possible predictive modes is then reduced to (9x 9x m) < (9x 9x 9). In the same way, if the mapping function f is used to flip the π sorted list, the predictive table can be further increased to 19 f $ 3B2 $ 7k 9¾ ία ilj ϊ Ε: Π steps are simplified. These methods can be used jointly or individually. ^ ^ ^ ...... as shown in Figure 7. Assume that the bribe is replaced by Negri using a certain internal pattern. The coding procedure of the bribe is as follows. The data blocks of the ordered coding frame are transmitted one by one to the encoder 5G of the view = conversion system shown in FIG. 7. The data block of the frame is converted from the image, and the input terminal 27 of the system receives digital image sources such as a camera or video recorder (not shown). In a known manner, the data block received from the digital image source includes the image pixel value. The 1 frame can be temporarily stored in the Λ frame β memory (not shown) or the encoder directly receives the data block input data one by one. . The lean blocks are transmitted one by one to the predictive method selection data block 35 to determine whether the pixel values of the existing data blocks to be coded can be predicted based on the coded data blocks in the previous item in the same frame or data segment. . To this end, the predictive method selects the data block 35 to receive input from the frame buffer of the encoder 33, including the previous item encoding and subsequent encoding and restructuring the records of the internal data blocks. In this way, the selection of data blocks by the foresight method can depend on whether the prediction of the existing data blocks can be performed based on the previous item of decoding and reorganizing the data blocks. In addition, if there is an appropriate decoded data block, when there is more than one method to choose from, the predictive method 35 may select the most appropriate method to predict the pixel values of the existing data block. Under certain circumstances, the prediction of the existing data blocks will not be carried out, because the appropriate data blocks used for the predictions do not exist in the frame buffer 20 factory i Ell punch β 33. When there is more than one prediction method, information about the selected prediction method will be provided to the multiplexer 13 for further transmission to the decoder. In some prediction methods, specific parameters required for prediction are transmitted to the decoder. Of course, it depends on the actual implementation method, and is not intended to limit the application of the mode coding in the present invention. The pixel values of the existing data blocks are foreshadowed in the internal predictive data blocks 34. The internal predictive data block 34 receives input about the selected predictive method from the predictive method selected data block 35, and obtains information from the frame buffer 33 about the existence data block applicable to the predictive. The predictive data block 34 can be used to make a prediction of an existing data block based on this information. The predicted pixel value of the existing data block is transmitted to the differential adder 28, and the predicted error data block is generated from the difference between the predicted pixel value of the existing data block and the actual pixel value of the existing data block received from the input 27. . Next, the error information of the predictive data block will be encoded into the predictive error coded data block in an efficient transmission form, such as using a discrete cosine transform (DCT). The encoding indicates that the error data block is transmitted to the multiplexer 13 for further transmission to the decoder. The encoder of the digital image transmission system also includes a decoding function. The encoding predictive error of the existing data block is decoded in the predictive error decoding data block 30 and then accumulated in the adder 31 together with the predictive pixel value of the existing data block. In this way, decoded versions of existing data blocks can be obtained. The decoded existing data blocks are transmitted to the frame buffer 33. It is also assumed here that the data blocks received by the receiver are formed into digital image frames one by one from the wheel channel. 21 i3®20m pages (\ w '. Cn μ. Αλ 30 ten ~ /] In the receiver 60, there is a multiplexer that receives the multiplexed encoding pre-error data block transmitted from the encoder 50. And predictive information. According to the predictive method, the predictive information may include the parameters used in the predictive procedure. If only one internal predictive method is used, no negative information about the predictive method is required for encoding the data block. In Figure 7 In the figure, the dashed lines are used to represent the selective transmission and reception of predictive method information and / or predictive parameters. Assuming that more than one internal predictive method can be used, information on the choice of the predictive method of the existing data block for decoding will be provided. The data block 41 is predicted in the data block 41. The data block 41 in the predictive data block will check the content of the frame buffer 39 to determine whether there is a previously decoded data block for predicting the pixel value of the existing data block. The image data block, which contains the predictive data block 41, will use the received predictive method information and the possibility of receiving from the encoder to indicate the predictive method shown by the M parameter to predict the existing data area. The content of the block. The predictive error information related to the existing data block is received by the predictive error decoding data block 36 and the appropriate method is used to decode the predictive error data block. For example, if it is encoded using the discrete cosine transform method The predictive error data =, the predictive error decoded data block will be reversed Da to search for the error, and then the error information is displayed in the predictor 1 and 37 of the existing image data block, and the output of the adder is supplied to the frame buffer When each data block is decoded, it will be guided to the solution surface only to crack the horse and accumulated in the frame buffer 39, and the shadow frame will be displayed.

The internal prediction data block 34 is a prediction of the existing data block based on the previous data block provided by the frame buffer 33 and previously decoded and subsequently decoded and reformed. In more detail, the prediction of the existing data block depends on the spatial prediction mode of the data block that was reformed before the use of the table in Table Gua or Table IV (not shown in Figure 7). However, when the sorting list of the predictive mode (i,] ·) of the lean material block currently being reorganized is omitted from the predictive table, the mapped data block 32 can be used to predict the space of the previously reorganized data block. The pattern is mapped to the complementary or mirrored space predictive mode (k, υ, where the internal predictive data block 34 can determine the complementary or mirrored predictive mode f (p) of the existing data block. The mapped data block 32 is again It is used to map the complementary prediction mode f (p) to obtain the predictive mode p of the existing data block. Similarly, the mapping data block 38 can be used for mapping when needed. It is used to perform (i, D to (k The mapping difference between the mapping of υ and the mapping operation to p can be coded by software program 69, the package. The machine executes the steps or virtual code by the method of this Maoming. The software program should be stored in a storage medium. For example For example, in the memory unit of TM or the one shown in Figure 8 = early 疋 68. Figure 8 shows a simplified schematic diagram of the child 90 used as a portable video communication device, in which the predictive mode of the present invention has been added. Display -I: Method. Active terminal 9 Equipped with at least-display module 76 for the whole = image capture device 7 2 'and-audio module 74 to obtain audio information from it and a key 7 «& audio frequency # signal of an audio production device. The movable terminal 90 should also include a disk 78 for inputting data and instructions,-RF component 64 for communication with 23 'and a signal / data processor 70 for controlling 60) Sun \' s operation. Best digital image data The block conversion system (50, bU) is implemented in the processor. According to another embodiment of the present invention, memory requirements can be limited and the field can violate coding efficiency. Various types of data block modes have been found through observation. Among the combinations, only a few of them have a high probability of occurrence. The other models have a low probability. In addition, the incidence of the remaining models will be similar, because the order of the M-type sorting set has little effect on the coding efficiency. The patterns specified by various combinations of predictive patterns can be divided into two groups. The first group includes m (where m is smaller than the overall number of existing patterns η) the most likely predictive patterns' and the first: group includes The rest of the patterns. The pattern of group- Sort according to their machine. As mentioned before, this sort is a list of models that are ranked from the highest probability to the lowest possibility. The models belonging to the second group can be sorted in a certain way. The month can be decoded according to the existing ones. The information of ^ is as follows: The small ^ predictive mode can be arranged before the large digitized prediction mode. In other words, the prediction mode is arranged into a sequential set s. The combination of the prediction mode specified in U and L The set in which the module is ordered is determined by the mode, T = {tl 't2, ...' tk} and V = {v 卜 v2, ..., vl}

VUmU R {tl, t], ..., tk, V], V2, ..., V!} _ The pattern of the -sorted set T is based on the expected combination of the pattern of the neighborhood data block and L Occurrence probability 24 is sorted. The higher the expected frequency of the pattern ', the fewer bits are required to provide a signal. The pattern sorting set is obtained by removing the order 5_first brother — 1 you-master fruit i and the order of retention / a, ·, and Wu. For example, in the sorting set s, as shown in i is in predictive mode Before j, 隼 人 V in the Chinese formula "before. The prediction should be in the predictive model, for example, 'considering the mode existing in the claw encoder and assuming that there is only the highest probability in each mode combination of the block U and L ^ In the first group, the remaining 8 patterns are listed according to the gradation of the number of them. ^: Is listed. To further illustrate this embodiment, consider the combination of f-form π and υ as the pattern with the highest probability. Pattern i constitutes a sorted set. T. The second sorting set v is formed by removing the pattern 1 of the sorting set s, that is, v = {0, 2, 3, 4, 5, 6, 7 '8} °. Here, the sorting set S is progressive by the number. The sequence of the nine predictive modes in the sequence is not formed, that is, S = {0,1,2,3,4,5,6,7,8}. Regarding whether the mode selected in a specific data block belongs to the first ° The information of the group or the second group is transmitted to the decoder. If it belongs to the first group, the transmitted information is displayed in the pattern group of the known data block ... Next, the 最高 th highest probability mode should be used for block c. In this embodiment, if the information received by the decoder shows that the first group is selected, since only the module belongs to this group, no further steps are needed. If the mode belongs to the second group, the transmission information shows the first mode in this group. In this case, 'if the received information _ should be the first mode, select mode 0. Example · 25 Same. The rules of ordering L in the set S may be different. 2. The elements of the first set τ are based on the adjacent data block u and the pre-pattern > $, ^. The elements of this set may be It is specified based on, for example, lists stored in the memory of the encoder and il R, or inferred by the modes u and L itself. The signal of a decoder such as .selection mode can be completed in several ways, for example-about the The rank of the predictive mode in which the sorting set is transmitted can be transmitted.--The information about the selected set or ¥ is based on the ranking of the selected set's mode. The preferred embodiment of the predictive mode specified in the JVT encoder is The description is as follows. Before specifying the predictive mode of data block C, the same code can be applied to different predictive modes of data blocks U and L. The diagonal modes 3, 5, and 8 are divided into groups and marked as 3 'and the diagonal mode [ 6, and 7 are divided into groups and marked as 4. After the grouping, each data block ML may have one of 5 types of beans labeled 0 '1, 2, 3, and 4. Therefore, from the original 9x9 The possible combinations of these predictive modes ML become only 5x 5 combinations. Nine predictive modes are arranged in a progressive manner according to their number, that is, s = (0 small 2, 3, 4, 5, 6, 7, 8), that is, Among the various combinations that can generate the ordered sets of predictive patterns U and L, only the 26 patterns with the highest probability are specified, that is, the ordered set τ of each combination has only one element. The designation methods are listed in Table V. Θ Table ν L / U outside 0 1 2 3 4 outside 0 0 1 0 0 0 1-0 〇〇1 0 1 0 l \ J 1 2 2 2 J 2 2 2 3 0 1 _2_ A 4 0 1 J > 4 4 The first information is sent to the decoder to decide whether to use the highest probability mode. This item is jointly coded to 4 × 4 data blocks, as shown in the table below.

Table VI

Codeword Block 1 Block 2 0 X X 10 110 X 111 X X represents the mode with the highest probability. In the data block that does not use the highest probability mode m, code numbers from 0 to 7 are transmitted to show which of the remaining modes should be used. Since the remaining patterns are ordered according to their number, 'the pattern with a smaller number will be ranked with the pattern with a larger number'. When the code number q is received, the predictive pattern is equal to · Q, if q < m 27 nmm, , 93. 10, -year / :! q + 1 'Otherwise, the code number q is transmitted as a 3-bit code, corresponding to the two-digit code number of the code number. ",, and L: Modular data: The memory saved by the two methods is derived from the data block, and the decoders of various combinations must rely on the most. At the same time, all modes of data blocks U and L Combinations:,; are the same. 7. b, ..., and a, the present invention provides a method, device, and device for encoding and decoding in bit stream image information. The image data block of the empty courtesy type indicated by the block. The spatial prediction pattern of the right block and the t block (C) can be based on the existing predictive pattern derived from the adjacent data block of the existing data block. Multiple decisions are grouped into the first group of "solid first predictive mode Γ and first; ΓΓ second predictive mode. The first prediction mode has a higher probability of appearing than the first prediction mode. The first-order is sorted according to its probability of occurrence, and the second pre-, first-, first-, and last- + 1 pre-order is sorted in a manner known to the decoder. Therefore, a decoder-sorting set will be provided. The t-decoder determines the second predictive mode ^ L order two set is the neighbor data block (u) of the existing data block (c). One or more groups of patterns are related. In addition, this bit is streamlined to show that when the derivative with the highest probability of occurrence is present: the encoding of the data block '' Which of the second pre-formula 2 has a targeted occurrence probability. Encoding information. The present invention also provides a computer program having a plurality of 28 derived predictive patterns grouped into virtual codes of the first and second groups, and the first predictive patterns can be sorted according to the occurrence probability of the first predictive patterns. The preferred embodiment of the present invention will be described, but those skilled in the art can make various changes, omissions and derivative forms without departing from the spirit of the present invention. 29 [Simplified description of the drawings] The first schematic diagram is as The 8 directional wedges used in the space prediction mode. Figure 2 is a schematic diagram showing the pixels of the existing 4 × 4 data block of the predicted pixel; Figure 3 is the two adjacent diagrams used for the prediction of the existing data block. Schematic diagram of the data block; Φ Figure 4a is a schematic diagram showing the spatial predictive mode of the two neighboring data blocks used by the * pre-existing f material block; Figure 4b is a mirror without the one shown in Figure 4a Schematic diagram of the space predictive mode of the adjacent data block according to the second relationship;-Figure 5a is a schematic diagram of another spatial predictive mode pair; Figure 5b is a schematic diagram of the mirrored mode pair, and Figure 6 is the flow of the spatial predictive method FIG. 7 is a block diagram of a digital image data block conversion system implementing a method of an embodiment of the present invention; FIG. 7 is a block diagram of a portable video communication device implementing a method of an embodiment of the present invention. Symbol description] 13 Multiplexer 21 Demultiplexer 27 Round-in terminal 28 Differential adder 29 Predict error compile device 30 30 Predict error decode data block 31, 37 Adder 32, 38 Map block 33, 39 Frame Buffer 34 indicates data block 35 indicates method selection data block 36 indicates error decoding block 40 decoder 41 indicates data block 50 encoder 60 receiver 64 radio frequency component 68 memory unit 69 software Program

70 CPU 72 Video capture device 74 Audio module 76 Display module 78 Keyboard 80 Audio generator 82 Input device 90 Active terminal 100 Method Step 110, 120, 130, 132, 140, 150 31

Claims (1)

Scope of patent application: [Scope of patent application] 1 · A method for encoding multiple bitstream image information composed of multiple image data blocks using multiple spatial predictive mode blocks using internal mode predictive data blocks, in which existing data areas The spatial predictive mode of block (c) depends on multiple derived predictive modes, which are derived from the spatial predictive modes of the adjacent data blocks (U, L) of the existing data block (c). The method is characterized by the following steps: Dividing the plurality of derived prediction modes into a first group and a second group, the first group having a first number of the first prediction mode, and the second group having a second prediction mode The second number, where the first number is less than the number of derived predictive modes, and the first predictive mode has a higher occurrence probability than the second predictive mode, and the predictive mode belongs to the first or second group of coded bitstreams Information to display the space of the existing data block (C) In addition, the following 2. The method described in item 1 of the scope of patent application, features: The probability of occurrence of the first predictive mode m2. The method described in item 1 of the patent scope, including the following predetermined method known to the detoxifier: the predictive mode of the row 32 1239207 sequence tube. History, 93.1 (Γ 4. · The method described in the first scope of the patent application, additionally including the following features: When the derivative predictive mode with the highest probability of occurrence is selected to encode 4nr (c), which of the first predictive modes will be displayed The most predictive mode of the most probable mode is coded into the bit stream information. The method described in Item 1 of the Litt HtT range additionally includes the following pre-sorted sorting sets that are known to the decoder, and then decode The device can determine the ordering of the two predictive modes according to the sorting set. The method of Leite 5nen additionally includes the blanking of the blanking block (U, L): the number of neighbors in the pre / data block (C) is 1 The above method is related to one or more groups. The method described in item 1 of the Litt range additionally includes the following number m model, which is represented by the number M, and is sorted according to the integers known to the decoder. 8. If the scope of patent application π ^ Man & What should be born in a derivative indicates the number of γ ten degrees, each integer is corresponding to the integer to, and who is too: "and the integer is sorted according to the method of providing a sorted set ^" The method additionally includes the following features: 33 The ordering set provided by the improved ordering set is applied to remove the integer corresponding to the / predictive mode, so as to provide a decoder to determine the ordering of the second group according to the ordering set. The method as described in item 1 of the patent application scope, wherein one of the plurality of derived pre-pulls is selected to encode an existing data block (C) '. The method further includes the following features: If the selected combination The predictive mode is in the second group and provides information about the ith mode of the calcite horse device in the second predictive mode. 10. The method as described in item 丨 of the scope of patent application, further including the features of column r: Mapping the spatial predictive mode of the neighboring data block to provide a complementary predictive mode of the neighboring data block when needed, The method is: ㈣❹ complementary predictive mode of data block to determine the complementary predictive mode of existing data block (c), and ~ mapping of the complementary predictive mode of existing data block ⑹ to obtain the space of existing data block (C) Prophetic mode. The method further includes the following features as described in item 10 of the scope of patent application: Complementary prediction of data block (C) 掇 々 η #The map of the under-preemption mode is ordered first and second by the mirror function Implemented by mirroring of data blocks. 34 1239207,94 ITT mode two uses the internal mode to predict the multiple spaces of the data block. Among them: two: the encoder of the image composed of the image data block, derived from the £ block (the space predictive mode of 0 depends on The spatial predictive pattern of multiple data blocks (U, L) is derived =, and the code has the following characteristics: = The multiple predictive patterns of the birth are divided into the first group and the second group: the first brother-the group has the first The first number of a predictive mode, and the second group: prepare the second number of the predictive mode, where the first number is less than the number of ^ :: twigs, and the first predictive mode has a greater problem than the second predictive mode. Probability of occurrence, and the second predictive mode is sorted according to a predetermined way of decoding equipment, and the median predictive mode that scarcely encodes bit stream information to display existing data blocks ((:) belongs to the first group or the second The device of the group is as follows: 13. The encoder according to item 12 of the scope of patent application, including the following features: ° The first predictive mode is sorted according to the occurrence probability of the first predictive mode to provide a sorted set. Device. 14. The encoder according to the application of patentable scope of item 12, further comprising the following features: G represents an offer, a decoder device information of the highest occurrence probability of pre-plant modes in the ordered set, 35. 15. The following features are included in the scope of the patent application: The encoder described in item 12 additionally includes a device that provides a complementary predictive mode when the field data block is required according to the space predictive mode of the field-based material block. Make the complementary predictive mode of the existing data block can be determined according to the complementarity of the adjacent data block, and provide the existing data block based on the mapping of the complementary predictive mode of the existing model (0 ^ Predictive mode device II. 16.—A method of decoding multiple bits of image data blocks composed of multiple image data blocks using multiple spatial predictive modes of the internal mode to predict data blocks, of which The space predictive model of the existing data block 于 is derived from multiple derived predictive models. The predictive model is derived from the spatial predictive models of the adjacent data blocks (U, L) of the Caibei material block (C). The method is characterized by the following steps: Wu " divides the multiple derived predictive patterns into a first group and a second group,-the group has the number of __ The second group has the second number of the second prediction mode, where the first number is less than the number of derived prediction modes, and the first prediction mode has a higher probability of appearance than the second prediction mode. This method is characterized by It lies in that—the space prediction mode from which the existing data block ⑹ is displayed is decoded by bit stream information belonging to the first or second group, so as to provide the decoder to determine the space prediction mode of the existing data block (C) according to the exhaustion information. Pirate 36 < The method described in item 16 of the Dodger Target, has the following characteristics: according to the order of the first set, including the pattern of grasses. ', Probability of the occurrence of the fruit, and the first-18, such as the following characteristics: Sequencing. The method described in item 16 of the scope of patent application, the method of knowing the predetermined way of the unlocking device includes a predictive model. • The method described in item 16 of the scope of patent application, below: Features. When selected When the derivative predictive mode with the highest occurrence probability is used to encode the existing data block (C), it will show which of the first predictive modes and the _predictive mode encoding person with the current probability The stream information includes 20. The method described in item 16 of the declared patent scope, further including the following features] & Ordering set for predictive mode to the decoder for the calcite horse to determine the second based on the ordering set Sorting of the predictive mode. 21 · —A decoder that decodes the bitstream image information composed of multiple image data blocks by using multiple spatial predictive modes of the internal mode predictive data block, in which the existing data block ( c) The space predictive mode depends on a plurality of derived predictive modes. The predictive mode is derived from the spatial predictive modes of the adjacent data blocks (U, L) of the existing data block (C), which is characterized by : The multiple predictive modes derived are divided into the first group and the second group, the first 37 L · Form-feed—a—j I The first number of the two predictive modes, and the second group has the Γ number of the second predictive number of projects, where the first number is less than the derivative predictive mode: base ': the first predictive The mode has a higher output than the second predictive mode. The feature of the 'δH' decoder is: a "space corresponding to the decoding information to depend on the existing data block (C). The pre-polarized system belongs to the first group or the second The device of the group, and the device for selecting the spatial prediction mode according to the above-mentioned decision. 22. The decoder according to item 21 of the scope of patent application, characterized in that the first prediction mode is based on the occurrence probability in the first prediction mode. ^ 2 3 · The decoder as described in claim 21 of the patent scope is characterized in that the second predictive mode is sorted in a predetermined manner, and the decoder additionally has the following characteristics: used to store the predetermined representation 24. The decoder according to item 21 of the scope of patent application, wherein when the derivative predictive mode with the highest occurrence probability is selected to encode the existing data block (C), The decisive device will be based on the decoded information to determine which one of the first predictive modes has the highest probability of occurrence. 25 · —A plurality of spatial predictive modes using the internal mode predictive data blocks are performed by multiple image data blocks. Image coding and resolution 38, Outline 7 5; coded image coding system, in which the spatial prediction mode of the existing data block (c) depends on multiple derived prediction modes, which are based on the existing data block (C ) Is derived from the spatial predictive patterns of multiple adjacent data blocks (u, L), which is characterized by dividing the multiple derived predictive patterns into a first group and a second group of devices, the first group having the first group The first number of predictive modes, and the second group has the second number of second predictive modes, where the __number is less than the number of derived predictive modes, and the first predictive mode has a higher probability of appearance than the second predictive mode A device that encodes the bitstream information representing the existing data block ⑹ into bitstream grouping information belonging to the first or second group, and uses the bitstream grouping information to A device for decoding for the decoder to determine the spatial prediction mode of the existing data block based on the decoded information. 26. The image editing system described in item 25 of the scope of patent application, and additionally includes the following features: The first group is a device that sorts the first predictive mode according to the occurrence probability of the first predictive mode. 27. The image coding system described in item 25 of the scope of patent application, further including the following features: Corresponding to the second group and The second predictive mode is arranged into a device known as a predetermined method of decoding. * ^ 39 rz 1239207 Pot 28 · —a computer program product, including a computer-readable medium, on which a group of executable codes is stored for use. In the image coding system, multiple spatial display modes are used to predict the data block in a more internal mode ~: 2: encoding of bit stream image information composed of image data blocks, 'space display of one block (C) The model is to determine the multiple derived models, factories, and models, and these predictive models are based on the space of the two neighboring data blocks (u, L) according to the data block (C) above. The predictive mode is obtained by shooting; the characteristic of the executable coding is that it contains: · the derived predictive mode is divided into the first group and the second group; -Number, and the second group has the second number of the second predictive mode, where the first number is + @ —, the number of the first predictive mode has a higher probability of occurrence than the second predictive non-nuclear type, and = Kangdi -The occurrence probability in the predictive mode is a virtual code used to sort the first predictive mode. α 29 is a computer program as described in item 28 of the scope of patent application. Or, the executable encoding further includes: a virtual code used for sequencing of the second predictive mode according to the pre-downtime mode of the M Bed + that the decoder is familiar with. 40 ifffiSl Revival d U L C Part 3 Flip Line Picture 4 匚. 1 s 6CO 6CNI 历 卜 濉 CNJ ml βίο, Μ ul- 一 il CVJ9 000 69 89 >